Lisong He

• Led experimentation across three Reels monetization workstreams, achieving incremental monetization gain. • Designed comprehensive A/B testing frameworks to guide product iterations and launch decisions. • Developed AI workflows for automated revenue diagnostics, reducing investigation time from days to hours.

• Enhanced auto insurance loss modeling by integrating housing-value features and retraining GBM models in H2O, achieving AUC lift. • Migrated legacy GBM models to XGBoost, improving claim severity prediction accuracy and model stability. • Developed a SAS-based fraud scoring system using logistic regression, identifying key risk factors for fraud management.

• Led a team of 4 in neurophysiological research, integrating 4000 sleep records using Pandas. • Conducted feature engineering to identify key attributes and relationships through Excel pivot tables. • Analyzed complex time-series sleep data with supervised learning algorithms, enhancing understanding of sleep patterns. • Achieved 91% accuracy with Support Vector Machine for pre-screening apnea cases, improving patient triage.

• Extracted, cleaned, and tokenized COVID-19-related tweets using Python’s nltk package. • Applied Latent Dirichlet Allocation with the “Sentiment140” dataset for sentiment analysis. • Trained Gradient Boosting Machines to classify sentiment and monitor shifts across virus variants. • Confirmed a negative correlation between public sentiment and death rates, revealing insights into public perception.